Matches in SemOpenAlex for { <https://semopenalex.org/work/W4205124414> ?p ?o ?g. }
Showing items 1 to 95 of
95
with 100 items per page.
- W4205124414 endingPage "4707" @default.
- W4205124414 startingPage "4694" @default.
- W4205124414 abstract "Deep neural networks (DNNs) achieved great cognitive performance at the expense of a considerable computation workload. To relieve the computational burden, many optimization works are developed to reduce the model redundancy by identifying and removing insignificant model components, such as weight sparsity and filter pruning methods. However, these works only evaluate model components’ static significance with parameter information, ignoring their dynamic interaction with external inputs. Specifically, due to the difference in per-input features, the model components’ significance can dynamically change and, thus, the static methods can only achieve suboptimal performance. Focusing on this aspect, we propose a dynamic DNN optimization framework in this work. Based on the neural network attention mechanism, we propose a comprehensive dynamic optimization framework, including 1) testing-phase dynamic feature map pruning; 2) training-phase optimization by training with targeted dropout; and 3) deployment-phase one-for-all (OFA) model adaptability enhancement. By providing a holistic dynamic testing, training, and deployment co-optimization framework, our work has the following benefits: first, it can accurately identify and aggressively remove per-input feature redundancy by considering the model-input interaction and involving the channel/column-wise pruning flexibility; meanwhile, the training-testing co-optimization favors the dynamic pruning and helps maintain the model accuracy even with a very high feature pruning ratio. Finally, the deployment enhancement provides one unified OFA model to support full-spectrum feature sparsity ratios. The unified model can be dynamically reconfigured to meet different resource budgets without any retraining cost, and thus provide significant deployment flexibility. Extensive experiments show that our method could bring 37.4%–54.5% floating-point operations reduction with negligible accuracy drop on various test benchmarks. Meanwhile, the OFA deployment optimization enables us to use one model to support at most ten different resource constraints without any retraining cost." @default.
- W4205124414 created "2022-01-25" @default.
- W4205124414 creator A5003870232 @default.
- W4205124414 creator A5025596795 @default.
- W4205124414 creator A5030027584 @default.
- W4205124414 creator A5032077042 @default.
- W4205124414 creator A5060314107 @default.
- W4205124414 creator A5076768386 @default.
- W4205124414 creator A5082386342 @default.
- W4205124414 date "2022-11-01" @default.
- W4205124414 modified "2023-10-17" @default.
- W4205124414 title "AntiDoteX: Attention-Based Dynamic Optimization for Neural Network Runtime Efficiency" @default.
- W4205124414 cites W2108598243 @default.
- W4205124414 cites W2143612262 @default.
- W4205124414 cites W2752782242 @default.
- W4205124414 cites W2792641098 @default.
- W4205124414 cites W2886851211 @default.
- W4205124414 cites W2889185260 @default.
- W4205124414 cites W2928560789 @default.
- W4205124414 cites W2963150697 @default.
- W4205124414 cites W2963363373 @default.
- W4205124414 cites W2964233199 @default.
- W4205124414 cites W2981698279 @default.
- W4205124414 cites W3012561096 @default.
- W4205124414 cites W3013407975 @default.
- W4205124414 cites W3028148258 @default.
- W4205124414 cites W3035467254 @default.
- W4205124414 cites W3035678286 @default.
- W4205124414 cites W4213084750 @default.
- W4205124414 cites W4236853429 @default.
- W4205124414 doi "https://doi.org/10.1109/tcad.2022.3144616" @default.
- W4205124414 hasPublicationYear "2022" @default.
- W4205124414 type Work @default.
- W4205124414 citedByCount "0" @default.
- W4205124414 crossrefType "journal-article" @default.
- W4205124414 hasAuthorship W4205124414A5003870232 @default.
- W4205124414 hasAuthorship W4205124414A5025596795 @default.
- W4205124414 hasAuthorship W4205124414A5030027584 @default.
- W4205124414 hasAuthorship W4205124414A5032077042 @default.
- W4205124414 hasAuthorship W4205124414A5060314107 @default.
- W4205124414 hasAuthorship W4205124414A5076768386 @default.
- W4205124414 hasAuthorship W4205124414A5082386342 @default.
- W4205124414 hasConcept C105339364 @default.
- W4205124414 hasConcept C105795698 @default.
- W4205124414 hasConcept C108010975 @default.
- W4205124414 hasConcept C111919701 @default.
- W4205124414 hasConcept C119857082 @default.
- W4205124414 hasConcept C138885662 @default.
- W4205124414 hasConcept C152124472 @default.
- W4205124414 hasConcept C154945302 @default.
- W4205124414 hasConcept C2776401178 @default.
- W4205124414 hasConcept C2780598303 @default.
- W4205124414 hasConcept C33923547 @default.
- W4205124414 hasConcept C41008148 @default.
- W4205124414 hasConcept C41895202 @default.
- W4205124414 hasConcept C50644808 @default.
- W4205124414 hasConcept C6557445 @default.
- W4205124414 hasConcept C86803240 @default.
- W4205124414 hasConceptScore W4205124414C105339364 @default.
- W4205124414 hasConceptScore W4205124414C105795698 @default.
- W4205124414 hasConceptScore W4205124414C108010975 @default.
- W4205124414 hasConceptScore W4205124414C111919701 @default.
- W4205124414 hasConceptScore W4205124414C119857082 @default.
- W4205124414 hasConceptScore W4205124414C138885662 @default.
- W4205124414 hasConceptScore W4205124414C152124472 @default.
- W4205124414 hasConceptScore W4205124414C154945302 @default.
- W4205124414 hasConceptScore W4205124414C2776401178 @default.
- W4205124414 hasConceptScore W4205124414C2780598303 @default.
- W4205124414 hasConceptScore W4205124414C33923547 @default.
- W4205124414 hasConceptScore W4205124414C41008148 @default.
- W4205124414 hasConceptScore W4205124414C41895202 @default.
- W4205124414 hasConceptScore W4205124414C50644808 @default.
- W4205124414 hasConceptScore W4205124414C6557445 @default.
- W4205124414 hasConceptScore W4205124414C86803240 @default.
- W4205124414 hasFunder F4320306076 @default.
- W4205124414 hasIssue "11" @default.
- W4205124414 hasLocation W42051244141 @default.
- W4205124414 hasOpenAccess W4205124414 @default.
- W4205124414 hasPrimaryLocation W42051244141 @default.
- W4205124414 hasRelatedWork W1538624230 @default.
- W4205124414 hasRelatedWork W2061300913 @default.
- W4205124414 hasRelatedWork W2961085424 @default.
- W4205124414 hasRelatedWork W3199608561 @default.
- W4205124414 hasRelatedWork W4286629047 @default.
- W4205124414 hasRelatedWork W4306321456 @default.
- W4205124414 hasRelatedWork W4306674287 @default.
- W4205124414 hasRelatedWork W4312263439 @default.
- W4205124414 hasRelatedWork W1629725936 @default.
- W4205124414 hasRelatedWork W4224009465 @default.
- W4205124414 hasVolume "41" @default.
- W4205124414 isParatext "false" @default.
- W4205124414 isRetracted "false" @default.
- W4205124414 workType "article" @default.